Autonomous Machine Patch Mapping for Unmapped Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomously moving machines face challenges in efficiently processing sensor data to navigate unmapped environments and identify obstacles, particularly with limited computing power and storage capacity, and require robustness against sensor failures.
Innovation Solution
The method involves generating a map frame comprising gravity patches from sensor data, where each patch represents a portion of the environment with assigned normals and depths, allowing for efficient image matching and projection, enabling the machine to operate with limited resources and navigate using a patch cloud that can be continuously updated.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the machine processes sensor data to navigate unmapped environments and identify obstacles, then the navigation capability is improved, but the computing power and storage capacity requirements increase
Solution Approach 1:
The patent segments the environment into discrete patches that can be independently processed and stored. Each patch represents a portion of the sensor data with associated depth information and normal vectors, allowing the system to handle large environments by dividing them into manageable units that can be processed with limited computing resources.
Solution Approach 2:
The patent creates a simplified graphical representation (copy) of the environment using patches that approximates the full sensor data. This patch-based model serves as a compressed representation that consumes less storage and computing power while maintaining sufficient accuracy for navigation and obstacle identification tasks.
2Measurement precision
If the machine stores detailed sensor data for accurate environment mapping, then the mapping precision is improved, but the storage capacity requirements increase
Solution Approach 1:
The patent divides the environment into patches, storing only essential information for each patch (position, normal vectors, depth values) rather than complete sensor data. This segmentation allows the system to maintain mapping precision for navigation purposes while significantly reducing the total storage requirements compared to storing full-resolution sensor data for the entire environment.
Solution Approach 2:
The patent applies different levels of detail to different regions by using patches with varying resolutions and detail levels. Areas requiring higher precision for navigation receive more detailed patch representations, while less critical areas use coarser representations, optimizing the balance between mapping precision and storage capacity.
3Adaptability or versatility
If the machine uses a patch cloud that can be continuously updated, then the adaptability to changing environments is improved, but the processing complexity increases
Solution Approach 1:
The patent implements a dynamic patch cloud that can be continuously updated as the machine moves through the environment. New patches are added, existing patches are modified, and outdated patches are removed based on the machine's current position and sensor inputs. This dynamic approach allows the system to adapt to changing environments while maintaining a manageable processing load by focusing computational resources on relevant patches.
Solution Approach 2:
The patent updates the patch cloud at periodic intervals based on machine movement and sensor data acquisition cycles. Rather than continuously processing all data, the system periodically integrates new sensor information into the patch model at structured intervals, reducing processing complexity while maintaining adaptability to environmental changes.
Data Source
AI summary
An autonomous machine and a method for operating the autonomous machine are disclosed. In an embodiment, the method includes receiving first sensor data from a first plurality of sensors supported by the machine, the first sensors covering a scene in a vicinity of the machine, generating a virtual map frame comprising a plurality of gravity patches and mapping the gravity patches and the first sensor data.